The Theory-Ladeness of Observation

The general answer to the question of authority presented by the empiricist consensus could be attacked on two different fronts: the evidential relationship claimed to hold between observation and theory and the status of the observational evidence. Philosophers working within the consensus were very much concerned with the former, and this concern issued in problems such as those involved in the technical development of inductive logic and the paradoxes of confirmation (the "gruesome ravens"). Additionally there were more or less constant skirmishes between inductivists, hypothetico-deductivists and Popperian falsificationists. However, the nature of the foundational "evidence" provided by observation was less of a concern to consensus defenders because they shared a commitment to empiricism which anchored authority firmly in sensory experience.

To be sure, among consensus defenders there was precious little agreement over just what it is that is "experienced" when an observation is made. In the earlier period of positivism, empiricists held out hope for a "sense data theory" of what it is which is observed; later they tended to move towards a rival view of observation known as "physicalism."
 

Sense data theorists argued that the observational foundation can ultimately be described in a purely perceptual vocabulary of "sense data" terms, such as colors and shapes, indexed with spatio-temporal terms like "here," "there," "then," and "now."
For these "hard empiricists" the world of "ordinary everyday objects," such as tables and chairs, people, animals, and plants, meter sticks, microscopes, and voltmeters are all "constructions" which could, in principle, be "reduced" to a vocabulary of purely perceptual terminology. Such a promised "reduction" proved immensely difficult (if not impossible) to actually achieve, thereby weakening the plausibility of this sense data approach.
The abandonment of sense data theory led other "softer" empiricists, such as the later Carnap, to embrace the doctrine known as "physicalism" which simply claimed that observation started with perception of the ordinary everyday objects the scientist deals with in assembling the observational evidence for a theory and no attempt is made to "reduce" them to bits of "sensory data." This view ultimately won out over sense data theory, which became less plausible after the Gestalt movement in empirical psychology made a good case for the claim that what are perceived are patterned wholes, not separable bits of data.
The insecurity over just what it is that is observed may have discomforted many consensus philosophers; nevertheless, they tended to present a united front on their conviction that whatever it is which the scientist observes, it is the foundation which justifies the choice of which theories gain acceptance into the domain of "scientific knowledge" and which supports the authority claimed by scientific explanations of the natural world.

One of the most influential of the attacks on the empiricist, foundationalist view of science shared by inductivists, hypothetico-deductivists, and falsificationists is the thesis that observation statements are "theory-laden." Because all of these views hold that the authority for the rational acceptability of scientific claims relies on a method by which theoretical statements are justified by observation, the attack on observation literally undermines the foundation of traditional views of science. Even if we grant that the observational evidence does confirm (or at least does not falsify, i.e. "corroborates") any given theory, the question still remains, does this confirmation by observation establish the rational acceptability of that theory?

In order to see why this is so, it is useful to reexpress the question that philosophers are addressing. Philosophers within the consensus often spoke as though the question facing the scientist can be expressed as "Does the observational evidence support this theory?" However, in real science the question is more usually in the form, "Which of two (or more) rival theories is better (or best) supported by the evidence?"

When we speak in this way of theories as being "rivals" we mean that a) they both purport to explain (at least some of) the same phenomena, and b) the explanations they offer are incompatible in the sense that they cannot both be considered "true" or that logically acceptance of one precludes acceptance of the other. Given this way of posing the question it becomes more obvious that we turn to observation to decide the choice between rival theories. Let us call this "the question of theory choice." In order for this decision to be rational, it is necessary that the observational evidence to which we turn is "neutral" at least with respect to these rivals. If it can be shown that this evidence by its very nature is such that it cannot be so "neutral," i.e. if the observational evidence is essentially theory-laden then appealing to it to justify the choice of one theory over another can no longer provide an objective ground on which to make such a rational choice.

No one would accept a verdict as "fair" if it were to be shown that all members of the jury were "friends" of the accused. Similarly, it can be argued that if one can show that all the evidence on the basis of which a theory choice is made is evidence which by its very nature is dictated by the very theory which is chosen, then the choice of that theory over its rivals can hardly be rationally justified. And this is precisely what the thesis of theory-ladeness claims.
Because we tend to underestimate just how radical it is, it will be helpful to be forewarned against several common misunderstandings of the thesis of theory-ladeness.
First it is not the thesis that an observer can see anything he or she wants to see. When I open my eyes, there is a real world presented to me which strongly constrains the possible experiences I can have; however, what I experience is not solely determined by the character of that world, but is to a significant extent also determined by my beliefs and expectations. When I report what I see, i.e. when I report an observation statement, that report is in turn strongly determined by the available concepts in the language of that statement.

Second, the theory ladeness objection to foundationalism is not equivalent to advocating the metaphysical idealist view that the world is a product of the activity of my, or any other, mind. Philosophers who hold the view that "reality" is through and through mental, that the world of experience is itself a product of the mind's activity, are called "metaphysical idealists." While this view was quite common in the previous century, it is not what is being claimed in the theory-ladeness attack on foundationalism. For those who advocate the theory-ladeness of observation, as for most philosophers of science, the mind-independent reality of "the world" is not in doubt. However, that mind-independent world, whatever its character might be, is not directly relevant to the question of the neutrality or theory-ladeness of observation. The observer reports what he or she experiences, and that is highly dependent on the character of the observer's thinking.

Third, the theory-ladeness thesis is not the uncontroversial claim that scientists often see what they want to see, i.e. what would confirm their theories, and ignore what they do not want to see, which would refute their theories. Of course scientists are human beings and are naturally enough prejudiced in favor of theories on which they have, perhaps, lavished great resources of time and money. Furthermore it is natural enough that scientists might tend to favor some theories over others for social, cultural, or religious reasons; in short not all theories are on an equal footing with respect to political or ideological correctness. All of these factors which may tend to tilt the question of theory choice one way or another may be referred to as "bias." The theory-ladeness claim is not the simple claim that scientists are often thus biased, for bias is something which a) we can be made aware of, and b) which we can correct for by exposing. Most scientists acting in good faith are aware of these facts and equally aware that their careers stand in grave jeopardy if it is demonstrated that they have injected biased observational reports into science. The sanctions against biased observation are sufficiently strong to eliminate much of this problem, but even in the cases where bias is so deep as to be more or less invisible to the person thus biased, there is the fact that scientific observations must be replicable. When scientists not sharing the original observer's bias attempt to reproduce his or her observational results, the effect of this bias will be discovered, and can thus, presumably be eliminated.

Finally, and most importantly, it must be grasped that those who argue for the theory-ladeness of observation are not making the rather uncontroversial claim that different observers see the same thing but interpret what they see differently. Few would dispute that. Even the most avid foudationalist admits that of course holders of rival theories interpret the "data" differently, each in the light of his or her own theory. However, a foundationalist claims that before these rival interpretations begin, so to speak, both observers are "given" the same datum to interpret (in Latin, "datum" simply means "given"). A sufficient base of data will eventually enable a neutral judge to determine which interpretation best fits the data. What the thesis of theory-ladeness is claiming is that observers with different beliefs "see" or "experience" different things, before any interpretive process can begin. Thus observation by itself can never settle disputes between such rival systems of belief. Since there is no common datum against which to measure the acceptability of rival theories, the observations made by holders of different theories are said to be "incommensurable."


The most famous expression of thesis of theory ladeness is found in the first chapter of N.R.Hanson's Patterns of Discovery, summarized below. Hanson is a tricky writer whose argument is not always obvious. The following summary from F. Suppe, The Structure of Scientific Theories, 2nd Ed., pp. 154-156 should help.

In his argument Hanson uses a method of "ordinary language analysis" pioneered by Ludwig Wittgenstein, which at the time Hanson wrote was very much a popular trend in philosophy in English speaking countries. Although this mode of argument is much less in favor these days, that fact does not substantially affect Hanson's argument. In order to follow Hanson's argument it will be convenient to distinguish between the verb "to see" and the verb "to look at." Hanson begins by pointing out that the claim that observers holding different theories experience different things is controversial only if there is another prior sense in which they experience the same thing. We will use "to see" to refer to what they experience which is different, and "to look at" to refer to what they experience which is the same.

Hanson considers a variety of cases where we encounter apparently differing observational reports under identical "given" conditions. Most famous of these is the (imaginary) case of the "geocentrist" Tycho and his assistant, the "heliocentrist" Kepler "observing" the sunrise. They experience the same thing in the sense of having equally normal perceptual organs which focus identical images of light on the retinas of their eyeballs. But do they "see" the same thing? We do not "see" retinal images, Hanson argues, that is not how the verb "to see," (what they called "seeing talk") is used. Retinal images of the sunrise are tiny, inverted, and two in number; none of which characterizes what we "see." Nevertheless Tycho sees the moving sun cross a stationary horizon, while Kepler sees a moving horizon dipping down to expose a stationary sun. So there is a difference in what they see.

Hanson now turns to counter an objection which is more or less the standar reply to this point: the difference arises in their "interpretations" of what they see. Let us call this the "interpretation thesis." According to this view there is in the visual experience a "two stage" event in which there is first, as it were, a "given" sensory experience which is identical for Tycho and Kepler, and then second, an "interpretation" put upon that given which, because of their different theoretical beliefs, differs in the case of our two astronomers.

Hanson opposes this "interpretation thesis" by considering a variety of simpler cases of drawings which can be seen as one thing or another, the most famous of which is perhaps the "pelican/antelope":

Viewed against one background it appears as a "pelican":

                    But against another the exact same "datum" appears as an "antelope
[Note the the image in the lower right of both the pelican picture and the antelope picture is "the same" as thge originbal image given above.]
 
Hanson argues first that if seeing the figure first as an antelope and then as a pelican "involves interpreting the lines differently in each case, then having a different interpretation ...just is for us to see something different. This does not mean we see the same thing and then interpret it differently." (p.9) The point Hanson is establishing here is that once it is admitted that interpretation does influence what you see, it must be granted that seeing the figure "under the pelican interpretation" and seeing it "under the antelope interpretation" amount to seeing two different things.

Secondly, he argues that the difference between seeing the pelican and seeing the antelope cannot be a matter of interpretation, because "interpreting" refers to a kind of thinking, an action of the mind, on something that is given to it to be interpreted, whereas seeing is an "experiential state." An interpretation is temporal, we can talk about being half way through an interpretation, but not in the case of moving from seeing a pelican to seeing an antelope. What are you doing as you read these words? Are you "interpreting marks on a page? When would that be a natural way of speaking?"

If the difference is not a matter of interpretation, then it must be something else; Hanson claims that since nothing optical changes, i.e., the retinal images are the same, the difference must lie in the organization of what one sees in the purely optical (as opposed to experiential) sense. But the "organization...is nothing that lies in the retina." What we experience when we say we "see the sun rise" will be organized differently in the case of those with different astronomical beliefs.

How one can organize what meets the eyeball depends on context and the perceiver's beliefs. When two observers looking at x see different things in x, that just amounts to having different beliefs or "theories" about x. Hanson thus introduces the word "theory-laden" which comes to refer to this whole argument: "There is a sense, then, in which seeing is a 'theory laden' undertaking." Observation is thus theory-laden in an essential way, which makes the "theory" an indispensable part of observation. (We cannot "regret" that it is so, for the very notion of observation without "theory" would make "seeing" irrelevant to theories. Yet, the empiricist consensus would claim, it is precisely for the sake a justifying the theories that we make the observation.)

Since the verb "seeing" is used in both "seeing as" and "seeing that" talk, Hanson argues that an essential part of seeing talk involves one's beliefs (or knowledge) about x. "Seeing" is not purely a perceptual event, but an "epistemic" event -a coming to hold certain beliefs, not simply a pre-epistemic passive perceptual receiving of some "given." "Seeing," at least insofar as it is relevant to science, is really best understood as "seeing that..." where what follows the "that" is a statement which has truth value.

Since "seeing" in science always involves "seeing that...," where what follows "that" is a statement, and since the statement that can be formulated to express the epistemic achievement expressed in the "seeing that..." (the coming to believe a certain belief about what one is seeing) is itself a function of the language of the observer, there is a "linguistic" element in "seeing," even thought there is nothing "linguistic" in what "meets the eyeball."

Hanson is working from a perspective already pioneered by Wittgenstein that different languages in effect express different world-views, or in the language of the specific sciences, different theories of the world. Rival theories about the world express their claims in different "languages" in which key terms (although perhaps expressed by the same "word," e.g. "sun") are given different meanings. Therefore those who hold rival theories will make different, and logically incompatible observation statements when "looking at" the same thing. Thus observation cannot settle the choice between rival theories, and thus cannot provide the "authority" to justify its acceptance over its rival.

Although Hanson, tragically, did not live to see effect that this argument would come to have on subsequent debate, it became a centerpiece of much argumentation in the decade (or two) of the "rationality crisis" which swept philosophy of science after the publication of Kuhn's Structure of Scientific Revolutions.